Automate Plea Forms via OCR

Develop a small prototype that takes a sample traffic citation (PDF/JPEG) and extracts key data fields (defendant name, citation number, date, violation, officer, and court) using OCR.
Focus on accuracy, structured JSON output, and error handling for low-confidence OCR results.

Deliverables:

Working OCR pipeline (Google Vision, Azure Read, or AWS Textract)

Extracted data displayed in a Google Sheet or JSON file

Confidence scores for each field

Documentation of parsing logic (regex, layout rules)

Note: Scanned images range from poor to good quality that are provided from customers.